01
Keep identity consistent
Use a reference image to guide a face, product shape, outfit, brand object, or visual identity.
MiniMax H3 R2V workflow
Use images, video clips, and audio references to guide how your MiniMax H3 video looks, moves, and sounds.
Reference-to-Video Studio
Combine a prompt with optional image, video, and audio references. Choose 768P or 2K output and generate up to 15 seconds.
Control what matters
Reference-to-video works best when every file has a job. Tell MiniMax H3 what to keep, what to borrow, and what to create from scratch.
01
Use a reference image to guide a face, product shape, outfit, brand object, or visual identity.
02
Use a short video clip to guide camera movement, action, pose, timing, or scene rhythm.
03
Use reference media to guide lighting, color, texture, composition, and mood.
04
Add audio context when you need the scene to follow a voice, ambience, music cue, or sound style.
R2V examples
Start from a real production goal. Each example shows the reference role and generation settings to try first.
Character reference
Use one character image and one motion clip to guide face, outfit, walk cycle, and camera feel.
Product reference
Use product photos to preserve shape and materials while adding camera movement and studio lighting.
Audio reference
Use audio context to guide pacing, mood, ambience, or a music-led visual sequence.
Motion reference
Use a reference clip for tracking, push-in, orbit, handheld motion, or action timing.
Style reference
Use visual references to keep color, texture, lighting, and composition aligned across a concept.
Frame control
Use an organized storyboard to guide the subject, shot order, framing, and ending across the video.
Prompt templates
A good R2V prompt tells the model what each reference should control. Use these as starting points.
How to use
Use fewer references with clearer roles. The prompt should explain the relationship between your assets and the video you want.
Upload images, video clips, or audio files that show the subject, motion, style, or sound you want to guide.
Tell the model which file controls identity, which file controls motion, and which file controls style or sound.
Start with 5 to 10 seconds for testing. Use 2K when detail matters, or 768P for a lower-credit draft.
Check faces, product details, logos, text, hands, audio timing, and camera motion before publishing.
Reference settings
The best setup depends on what you need to control. Start simple, then add more context only when it improves the result.
Goal
Best reference
Prompt instruction
Suggested start
GoalCharacter identity
Best referenceClear portrait or full-body image
Prompt instructionKeep face, hair, outfit, and body type from the image.
Suggested start10s · 16:9 or 9:16
GoalProduct accuracy
Best referenceProduct photo from clean angles
Prompt instructionPreserve shape, material, logo area, and color.
Suggested start6–10s · 2K
GoalCamera movement
Best referenceShort motion video
Prompt instructionBorrow only camera path and pacing from the video.
Suggested start5–8s · 768P draft
GoalAudio mood
Best referenceMusic, ambience, or voice reference
Prompt instructionFollow the tone and rhythm without changing the scene brief.
Suggested start10–15s · image or video also required
Choose the right input
Use the reference type that matches the control you need. Adding more files is not always better.
Best for identity, product shape, outfit, color, visual style, or composition.
JPG, JPEG, PNG, WEBP, HEIC, or HEIF · up to 9
Best for motion, camera path, pose, transition, timing, or action rhythm.
MP4 or MOV · up to 3 · 15 seconds total
Best for mood, voice direction, music-led visuals, ambience, or sound design.
WAV or MP3 · up to 3 · 15 seconds total
Reference audio cannot be used by itself. Include at least one reference image or reference video.
Use cases
Reference-to-video is useful when the video needs to follow existing material instead of starting from a blank prompt.
Turn product photos into short ads, launch clips, and campaign concepts.
Use a character image to guide identity across a short story or motion test.
Create vertical concepts for Reels, Shorts, TikTok, and paid social creative.
Use audio direction to shape mood, pacing, rhythm, and ambience.
Bring character art, environment concepts, and trailer ideas into motion.
Test shot direction, camera movement, lighting, and scene rhythm before production.
Review before publishing
Check faces, hands, product geometry, packaging text, logo placement, brand colors, voice quality, and audio timing before using a generated video in a live campaign. For important brand assets, generate multiple drafts and keep the version with the cleanest details.
Credits
Reference workflows can need more iteration because you are guiding several details at once. Start with a shorter clip, then move to a longer or higher-resolution output when the direction is right.
View credit packs →Credit rate
768P · 9 credits/sec
2K · 14 credits/sec
For Reference-to-Video, billable time equals reference video duration plus output video duration. Images and audio do not add billable seconds.
FAQ
It is a reference-led generation workflow. You add images, video clips, or audio so MiniMax H3 can use them as context for subject identity, motion, style, camera direction, or sound.
Yes. Images are useful for character identity, product shape, outfit, visual style, and composition.
Yes. Video references are useful for camera movement, pacing, action timing, pose, and motion direction.
Yes. Audio references can guide mood, voice direction, music, ambience, or sound style. Audio cannot be used alone; include at least one reference image or video.
Image-to-video starts from a required first frame and can include an optional last frame. Reference-to-video can combine images, videos, and audio to guide different parts of the generated clip. The two modes cannot be mixed in one request.
You can create an output up to 15 seconds long. Reference video clips can total up to 15 seconds.
Yes. Choose either 768P or 2K before generating.
Use the fewest references that clearly explain your goal. You can add up to 9 images, 3 videos, and 3 audio files, with no more than 12 materials in total.
Start creating
Upload images, video clips, or audio. Write a clear prompt. Generate a reference-led video in one workflow.